As a natural consequence of their energy consumption, technology companies are increasingly fitting the mold of large industrial energy consumers, signing power purchase agreements (PPAs) to ensure a secure and continuous supply of energy for their operations. Amazon, for example, has become the world's largest corporate buyer of renewable energy, signing more than 500 PPAs across 27 countries – a tally on par with some European nations.

To fuel the rapidly expanding US data center sector, technology leaders such as Amazon, Google and Microsoft are actively pursuing a secure and sustainable power supply. With the US hosting more than 50 gigawatts (GW) of data center capacity in 2024, tech companies are leaving no stone unturned as they evaluate a range of options – from solar photovoltaic (PV) and battery storage, to gas and nuclear power. While renewables and batteries are advancing, small modular reactor (SMRs) technologies, which could provide baseload supply and flexibility, still need to prove their commercial viability. Both solutions present merit, but one thing is clear: the US data center boom is coming, and it urgently needs power by any means necessary.

This also extends beyond the US. According to Rystad Energy’s research, global data center electricity consumption is projected to more than double by 2030. By 2040, power demand could soar to 1,800 TWh – enough to power about 150 million US homes for a year – as major tech companies continue to expand their processing capacity.

While this ramp-up may seem threatening, the sizeable energy demands of data centers – if managed effectively – could help stabilize local power grids.

To evaluate the stress on power grids at a granular level, it is crucial to understand the intricacies of training AI models and their batch-processing nature. These models gather and process data infrequently, allowing data centers to manage their energy use effectively. This is done through power-capping, which limits the maximum power that processing units can consume and reduces energy consumption, while only marginally raising the time taken to complete tasks.

Meanwhile, AI model training can be paused and resumed to support energy-efficient scheduling, which can be short-term or long-term. Short-term scheduling shifts workloads to times when renewable energy sources – such as solar power during the day – are plentiful and power prices are low. Long-term scheduling involves planning for different seasons, running more processes in the summer when energy costs are lower, and scaling back in the winter when prices rise. These strategies can also be adapted in real-time to optimize power use by moving workloads to off-peak hours, helping to balance the energy grid.

With this in mind, big tech companies are searching globally for suitable locations to build energy-intensive data centers, and countries such as Norway stand out as ideal candidates. Norway has historically offered low power prices, along with a high share of clean hydropower and a cold climate that naturally cools the heat generated by data centers.

Norway's ability to provide flexible energy to the European grid is becoming increasingly important, especially as the transmission grid struggles to meet high demand during peak times. Many people and political groups are concerned that data centers could drive up electricity prices for households during these times. However, with appropriate regulations, data centers could use energy flexibly, acting as reliable buyers for power producers when there is excess supply. This elastic demand could help stabilize overall consumption, optimize grid utilization, and reduce price volatility.

However, increasing interest in Norway as a data center hub has ignited political debate, highlighting the need for discussions on how to move forward. Proposals that have surfaced include a licensing system that incorporates criteria for social benefit and the use of waste heat, along with potential measures to limit data storage.

Globally, the data center boom has sparked mixed responses. Ireland and the Netherlands are restricting new developments due to grid strain and other concerns, while other countries are seeking solutions to accommodate surging electricity demand. Singapore, for instance – recognizing the economic potential of data centers – has lifted previous restrictions and is exploring changes in legislation to accommodate the data center boom, in a bid to ensure that the city-state does not fall behind.

The willingness to explore unconventional options highlights the efforts nations are taking to capitalize on the data center market while attempting to address the inherent challenges, particularly regarding energy supply and their alignment with decarbonization goals.

Market liberalization adds another checkbox for tech giants, as current regulatory bottlenecks could hinder their future ambitions. Thailand is one country that fits the bill. While its grid and climate policies differ from Norway’s, the Southeast Asian nation is actively pushing for power sector deregulation.

By opening its market to competition and loosening government control, Thailand is creating a more attractive environment for private investment. This proactive approach has already garnered significant interest, including 47 data center projects that have raked in more than $5 billion in investments as of December 2024.

As data centers multiply, managing their electricity demand and expansion requires a patchwork of regulatory and development strategies. While this balancing act is essential, sound policy and infrastructure investments can play a contributing role. Strategic measures, including flexible energy consumption, real-time demand response and the integration of renewable power sources, can help alleviate these pressures.

Discover more about our capabilities in the power sector through our Renewables & Power Solution, here.